Re: [PATCH] perf unwind-libdw: Fix reading the stack of a 32-bit task

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [Low] perf: incorrect data extraction for unaligned stack reads during cross-endian unwinding
--

commit c969a9ae6966b9f317efd438b881a5fa16635c84
Author: Stian Halseth <stian@itx.no>

perf unwind-libdw: Fix reading the stack of a 32-bit task

This commit fixes reading the stack of a 32-bit task during unwinding on
big-endian machines by reading 4 bytes instead of 8, preventing traps. It
also corrects word ordering for cross-endian unwinding.

> diff --git a/tools/perf/util/unwind-libdw.c b/tools/perf/util/unwind-libdw.c
> index 4ccfcc7c2dfca..d49901b1571e9 100644
> --- a/tools/perf/util/unwind-libdw.c
> +++ b/tools/perf/util/unwind-libdw.c
[ ... ]
> +/*
> + * libdw expects a 32-bit task's words zero-extended.  A recording of the
> + * other byte order was swapped in 8-byte units when it was read (see
> + * perf_event__all64_swap()), so a 4-byte word is picked out of its unit with
> + * that swap undone, as for PERF_SAMPLE_CPU in __evsel__parse_sample().
> + */
> +static Dwarf_Word stack_word(struct stack_dump *stack, int offset, size_t len,
> +			     bool swapped)
> +{
> +	union u64_swap u;
> +
> +	if (len == sizeof(u64))
> +		return *(Dwarf_Word *)&stack->data[offset];
> +	if (!swapped)
> +		return *(u32 *)&stack->data[offset];
> +
> +	u.val64 = bswap_64(*(u64 *)&stack->data[offset & ~7]);

[Severity: Low]
Does this bitwise logic truncate the offset to a 4-byte boundary?

If a DWARF CFI expression requests a stack read at an unaligned address
during cross-endian unwinding, the calculation discards the unaligned bits.

While register unwinding typically targets 4-byte aligned addresses on
32-bit architectures, could a malformed or custom ELF with non-standard
CFI trigger an unaligned read? If so, would this extract the wrong bytes
instead of the requested data, potentially corrupting the unwind results?

> +	return bswap_32(u.val32[(offset & 4) / 4]);
> +}

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260923193432.2489729-1-stian@itx.no?part=1